Comparison of Machine Learning Algorithms to Predict Cardiovascular Heart Disease Risk Level

Abstract

Cardiovascular diseases can pose a potential risk for almost every individual since they are associated with multiple parameters such as chronic disease, lifestyle, especially genetic factors. For this purpose, within the scope of the study, machine learning-based models were developed to predict the cardiovascular disease risk level and the metric performances of the algorithms were compared. For this purpose, the performances of the algorithms of the models developed using a data set accessible to all researchers were analyzed in a versatile way. In the study, the results obtained using Logistic Regression, Decision Trees, Random Forests, K-Nearest Neighbors, Gaussian Naive Bayes and LightGBM algorithms were compared. The results present the performance of each algorithm by evaluating it on metrics such as accuracy, precision, sensitivity and F1 score. The study aims to illuminate in which situations different algorithms are more effective and which variables are more determinant in terms of risk estimation. The results of this study can be used as an auxiliary diagnostic method for healthcare professionals working in the cardiovascular field. It can also be used as a predictive model for individuals who want to use artificial intelligence to determine the level of risk.